def score_chunks_tertiary(X, id_num):
"""
Score a single event based on:
Features: body, subject and event
Trained secondary classifier
Outputs highest scoring predicted primary classification and score
"""
X = secondary_multiple_events.score_chunks_secondary(X, id_num)
# load classifier (stored as list of classifiers)
clf_list, clf_names = ml_utils.load_classifier_list(id_num, 'tertiary')
# load feature pipeline
feature_pipeline = ml_utils.load_classifier(FEATURE_DIR + "tertiary_" +
str(id_num))
# transform data using feature pipeline
x_test_features = tertiary_train_model.get_tertiary_testing_features(
X, feature_pipeline)
results, classes = ml_utils.get_classifier_results(clf_list, clf_names,
x_test_features, None)
tertiary_pred, tertiary_score = tertiary_train_model.get_ensemble_prediction(
results, classes)
X['predicted_tertiary_class'] = tertiary_pred
X['predicted_tertiary_score'] = tertiary_score
X = tertiary_train_model.add_final_tertiary_classes(X)
return X
def main():
"""
Main function that kicks off the program
"""
start_time = time.time()
logger.info("Training new tertiary model...")
args = parser.parse_args()
id_num = args.id
df = pd.read_pickle(OUTPUT_DIR + "_secondary_dataset_" + str(id_num) +
".pkl")
y = df['true_tertiary_class']
X = df[[
'subject', 'text', 'event_type', 'final_primary', 'final_secondary',
'predicted_primary_score', 'predicted_secondary_score'
]]
logger.info("Size of X input: %s", X.shape)
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.4)
x_train_matrix, feature_pipeline = get_tertiary_training_features(
x_train, y_train, TERTIARY_FEATURE_PIPELINE)
x_test_matrix = get_tertiary_testing_features(x_test, feature_pipeline)
save_feature_pipeline(feature_pipeline, 'tertiary', id_num)
clf_list, clf_names = grid_search(x_train_matrix, y_train,
CLASSIFIER_PIPELINE)
results, classes = get_classifier_results(clf_list, clf_names,
x_test_matrix, y_test)
save_classifier(clf_list, 'tertiary', id_num)
y_pred, y_score = get_ensemble_prediction(results, classes)
_check_prediction_dimensions(y_test, y_pred, y_score)
x_test = _decorate_with_tertiary(x_test, y_test, y_pred, y_score)
x_test = add_final_tertiary_classes(x_test)
# Calculate the accuracy of the model
logger.info("------------------------------------------------")
logger.info(
"Overall Accuracy Tertiary: %s",
accuracy_score(x_test.true_tertiary_class, x_test.final_tertiary))
print(
precision_recall_matrix(x_test.true_tertiary_class,
x_test.final_tertiary, classes))
logger.info("Storing Tertiary model output and saving...")
# saving true tertiary classes
save_dataset(x_test, 'tertiary', id_num)
logger.info('Completed training of tertiary model %s', id_num)
logger.debug("Elapsed time: %s", time.time() - start_time)
def main():
"""
Main function that kicks off training
"""
start_time = time.time()
logger.info("Training new primary model...")
X, y = get_primary_inputs()
x_train, x_test, y_train, y_test, _, idx2 = train_test_split(X,
y.p_class,
X.index,
test_size=0.4)
# saving for future models
y_secondary = y.s_class[idx2]
y_tertiary = y.t_class[idx2]
x_train_matrix, feature_pipeline = get_primary_training_features(
x_train, y_train, PRIMARY_FEATURE_PIPELINE)
x_test_matrix = get_primary_testing_features(x_test, feature_pipeline)
id_num = save_feature_pipeline(feature_pipeline, 'primary', None)
clf_list, clf_names = grid_search(x_train_matrix, y_train,
CLASSIFIER_PIPELINE)
results, classes = get_classifier_results(clf_list, clf_names,
x_test_matrix, y_test)
save_classifier(clf_list, 'primary', id_num)
y_pred, y_score = get_ensemble_prediction(results, classes)
_check_prediction_dimensions(y_test, y_pred, y_score)
# Calculate the accuracy of the model
logger.info("------------------------------------------------")
logger.info("Overall Accuracy Primary: %s", accuracy_score(y_test, y_pred))
print(precision_recall_matrix(y_test, y_pred, classes))
logger.info("Storing Primary model output and saving...")
# saving true secondary classes
x_test = decorate_with_primary(x_test, y_test, y_pred, y_score,
y_secondary, y_tertiary)
save_dataset(x_test, 'primary', id_num)
logger.info('Completed training of primary model %s', id_num)
logger.debug("Elapsed time: %s", time.time() - start_time)
def score_chunks_primary(X, id_num):
"""
Score a single event based on:
Features: body, subject and event
Trained primary classifier
Outputs highest scoring predicted primary classification and score
"""
# load classifier (stored as list of classifiers)
clf_list, clf_names = ml_utils.load_classifier_list(id_num, 'primary')
# load feature pipeline
feature_pipeline = ml_utils.load_classifier(FEATURE_DIR + "primary_" +
str(id_num))
# transform data using feature pipeline
x_test_features = primary_train_model.get_primary_testing_features(
X, feature_pipeline)
results, classes = ml_utils.get_classifier_results(clf_list, clf_names,
x_test_features, None)
y_pred, y_score = primary_train_model.get_ensemble_prediction(
results, classes)
return y_pred, y_score